Neuron Explorer

Qwen3-0.6B

word before 'and' or 'or' in pairs

A neuron in layer 15 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word that sets up a pair, like 'both good' in 'both good times and bad times' or 'either in' in 'either in person or online', where the text is about to say 'and' or 'or'. When it was turned up, the model fell into long chains of 'and the...and the...', so it seems to push the model toward joining things with 'and'.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    56-1930) and Frank Meadow Sutcliffe⏎(1853-1941). The show divided the critics into those that saw that photography was indeed an art and those

  2. #2 peak

    firm goal in mind motivates them to move ahead on a project. Goals are most useful when they are specific, realistic, and important to you. Consider both short- and long-term goals. Your success depends on setting goals that really matter to

  3. #3 peak

    , changes in treatment guidelines, progress with similar initiatives that might shape attitudes of policy makers towards your issue, what has happened with similar initiatives in other areas (both successes and failures), and shifts in political priorities and policy intentions. Keeping a diary of

  4. #4 peak

    be requested to compare sales reports from two different models. Similarly, if you’re analyzing the affects of smoking on children’s asthma, you can compare youth asthma symptoms before and after smoking cessation. Such

  5. #5 peak

    is a section on the connections between the NGSS and the course topics.⏎I could see this being used for independent study or by a study group (either in-person or online). And it’s FREE.<|endoftext|>October 30, 2

  6. #6 peak

    skills and he was a good listener.⏎Over a period of 240 years there has been 44 presidents and the U.S has had both good times and bad times and that all depended on who was running the country and what choices

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: the, I, have, a, a, A, A, A, A, A, A, A, A, A, A, A, A, A, A, A, A, A, A, A,

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    . Article LA Times.<|endoftext|>Applied Behavior Analysis (ABA) therapy is an evidence-based best practice treatment that focuses on the science of behavior and learning to overall decrease negative behaviors and increase positive behaviors. ABA therapy utilizes the understanding of how behavior works

  2. #2 peak

    a formal account at a financial institution, according to the World Bank’s Global Financial Inclusion Database. As a result, most poor households operate almost entirely in the cash economy, particularly in the developing world. This means they use cash, physical assets

  3. #3 peak

    folks who have high blood pressure, even more evidence points to the benefits of the DASH diet (an eating plan that encourages a lot of fruits, vegetables, whole grains, and heart-healthy fats). Since these diseases and conditions are all too

  4. #4 peak

    in a greenhouse or under cloches e.g. tomatoes, peppers, courgettes are planted out after being ‘hardened-off’ (kept out during the day and in at night for a week or so) and after<|endoftext|>The Prolet

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L14:1830 excites
  2. L14:1718 excites
  3. L13:737 excites
  4. L14:229 excites
  5. L14:2089 excites
  6. L14:864 suppresses
  7. L11:1387 suppresses
  8. L12:2253 suppresses
  9. L10:1115 suppresses
  10. L13:1981 suppresses
  11. L11:2002 excites
  12. L14:2033 excites
  13. L14:2969 suppresses
  14. L12:537 excites
  15. L11:1511 excites
  16. L14:779 suppresses 'than' in comparisons
  17. L14:2986 excites
  18. L13:20 suppresses
  19. L12:2485 suppresses
  20. L14:1045 suppresses
  21. L14:1841 excites
  22. L14:2732 excites
  23. L14:2203 excites
  24. L8:2154 excites

Feeds later layers

  1. L18:120 strongly suppresses
  2. L18:2278 excites
  3. L18:104 excites
  4. L19:1605 excites
  5. L18:957 excites
  6. L24:205 excites
  7. L19:2308 excites
  8. L20:852 suppresses
  9. L17:1036 excites
  10. L16:565 suppresses
  11. L18:1207 excites
  12. L16:473 suppresses
  13. L18:2346 excites
  14. L20:1009 excites
  15. L16:2992 excites
  16. L19:2835 excites
  17. L17:2610 excites and or comma closing a list
  18. L19:164 suppresses
  19. L19:1092 excites
  20. L17:632 excites unclear: coordinating words in lists
  21. L19:584 suppresses
  22. L25:6 suppresses
  23. L19:1394 excites
  24. L19:628 excites

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -3.42 and -3.311 values between -2.86 and -2.751 values between -2.07 and -1.961 values between -1.74 and -1.622 values between -1.62 and -1.511 values between -1.51 and -1.402 values between -1.40 and -1.291 values between -1.29 and -1.172 values between -1.17 and -1.066 values between -1.06 and -0.955 values between -0.95 and -0.845 values between -0.84 and -0.735 values between -0.73 and -0.6117 values between -0.61 and -0.5027 values between -0.50 and -0.3971 values between -0.39 and -0.28155 values between -0.28 and -0.16466 values between -0.16 and -0.051266 values between -0.05 and 0.06927 values between 0.06 and 0.17533 values between 0.17 and 0.29251 values between 0.29 and 0.4099 values between 0.40 and 0.5140 values between 0.51 and 0.6215 values between 0.62 and 0.743 values between 0.74 and 0.852 values between 0.85 and 0.963 values between 0.96 and 1.07offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. 和其他
  2. 和
  3. and
  4. 和地区
  5. versus
  6. 和
  7. 与
  8. 以及其他
  9. 及其他
  10. 以及

Suppresses

  1. ща
  2. cla
  3. cle
  4. казал
  5. ifier
  6. Kİ
  7. лу
  8. CLA
  9. atra
  10. 他就